Unlocking Secure Collaboration Between Platforms and Advertisers

‍

Data-driven advertising thrives on insights from user behaviour, but concerns over privacy often limit what can be shared between platforms and advertisers. In today’s world, users demand transparency and control over their data, and regulations like GDPR and CCPA enforce strict privacy standards. Vaultree’s VENumML (Vaultree Encrypted Numbers Machine Learning) offers a revolutionary solution, enabling secure collaboration that protects user data while delivering actionable insights.

In the Social Media Demo, VENumML illustrates how a social media platform and an advertiser can collaborate securely to identify and target potential customers without compromising privacy. Let’s dive into the details.

‍

A Real-World Use Case: Secure Advertising on Social Media

‍

Imagine a social media platform, ClickReadShare, managing private subforums where users discuss sensitive topics. One such subforum, “Health and Wellness,” hosts conversations on sleep health. An advertiser, Snoo-Ze-Time, seeks to promote its sleep tracking app to relevant and influential users within this subforum. However, privacy concerns prevent ClickReadShare from sharing raw user data.

The Problem: How can ClickReadShare and Snoo-Ze-Time collaborate to identify potential customers while ensuring user data remains confidential?

The Solution: Using VENumML, ClickReadShare encrypts user data, including a graph of user connections and summaries of their posts, before sharing it with Snoo-Ze-Time. The advertiser uses this encrypted data to analyse user behaviour and rank them based on relevance and influence. Encrypted results are sent back to ClickReadShare, where users are identified for targeted advertisements—all without exposing sensitive information.

‍

‍
Breaking Down the Demo

‍

Let’s delve into the technical steps of the demo, which combines graph exploration and state-of-the-art natural language processing (NLP) with advanced encryption techniques:

1. Encrypting the Data: ClickReadShare encrypts two key datasets:
   • User Summaries: These include anonymised descriptions of user activity in the “Health and Wellness” subforum. For instance, posts discussing sleep health or fitness routines.
   • User Graph: A network graph showing who follows whom in the subforum, representing each user’s influence within the community.
   • This encryption ensures that sensitive information, such as personal health details or network relationships, is never exposed.‍
2. Transforming Data with Encrypted Machine Learning: Snoo-Ze-Time processes the encrypted data using VENumML’s advanced tools:
   • User Classification: A transformer-based model, trained on encrypted data, classifies user summaries to identify those with a high interest in sleep health.
   • User Ranking: PageRank, an algorithm traditionally used to rank web pages, is applied to the encrypted user graph. This helps identify influential users who are likely to amplify Snoo-Ze-Time’s message.
   • All computations are performed on encrypted data, maintaining user privacy throughout the process.‍
3. Sharing Secure Insights: Once the analysis is complete, Snoo-Ze-Time sends encrypted results back to ClickReadShare. These results include:
   • Interest Scores: How relevant each user is to Snoo-Ze-Time’s product.
   • Influence Rankings: A score indicating each user’s ability to influence others within the subforum.
   • ClickReadShare decrypts this data to identify users for targeted advertisements.

‍

Key Features Unlocked by VENumML

‍

The demo highlights VENumML’s ability to address critical challenges in privacy-preserving analytics:

1. End-to-End Encryption: User data remains encrypted during preprocessing, analysis, and sharing, eliminating risks of exposure.
2. Advanced ML on Encrypted Data: Supports transformer-based models for complex text analysis and graph algorithms like PageRank for ranking.
3. Scalability: Handles large datasets, typical of real-world social media platforms.
4. Compliance with Privacy Regulations: Meets the requirements of GDPR, CCPA, and similar frameworks.

And this is more than a technical showcase—it’s a blueprint for privacy-preserving collaboration. By enabling secure data sharing, VENumML empowers:

• Social Media Platforms: To monetise their data responsibly without breaching user trust.
• Advertisers: To target relevant customers without accessing sensitive information.
• Users: To benefit from personalised experiences without compromising their privacy.

‍

Get Hands-On with VENumML

‍

Curious to try VENumML? Meet us on Github to try it out! This demo notebook, as well as others, will walk you through the steps of building and deploying an encrypted transformer. It’s designed for developers and researchers who want to explore the cutting edge of privacy-preserving AI. If that sounds like you, we’d love to hear from you! If there is anything you need, let us know via Github issues or our Support portal.

‍